Geez....this is such a technical thread that I was nervous about posting, but I will. This post in a thread on Gearslutz revived an old thread and spoke to my inner nerd. If you ever cared about real synths this is so fascinating!!!
…Original question…
Barfunkel
Lives for gear
Thread Starter
Why can't they just make more chips?
Disclaimer: I know absolutely nothing about electronics or how they are made, so forgive me if this is a stupid question.
There's lots of classic chips no longer made. Tons of synths, even modern ones, that can't be manufactured or repaired because replacement chips simply don't exist anymore.
So... what's stopping anyone from making them? Is it horribly expensive to make a small run of a 1000 or so of CEM's or such? Do they use materials that are impossible to get? Are the schematics heavily guarded secrets?
I mean, this is 2010. We got flying cars and lasers and everything, why is a 30 year old chip impossible to recreate?
Even if they were expensive to make, why wouldn't someone make a small patch and ask a ridiculous price? Someone probably would buy them, right? People pay insane prices for the synthesizers themselves, it's not so far out to assume people would also pay insane prices to get them fixed, build new ones or build new, similar synths. There's probably at least 10k guys who would buy an SP1200 the second it was available again, almost regardless of the price. I know I would.
Reply :
Guys, this is a most interesting thread to me, and yes, the government certainly does run into these issues.
I have been knee deep in the Curtis designs for a few months now. I work with a couple incorrigible nerds, electrical engineering geeks who think "phonesex" is talking to Texas Instruments technical documentation librarians and engineers about things called "teensies"... so as a defense against their constantly derailing good conversations about hookers and cigars with "Arduino" suggestions for the water-cooler temperature, I have taken picking their brains on solves and prior cases where obsolete semiconductors have been 'updated' in prototype and other oddball use-specific cases.
I already knew this but my captive help confirmed it between laughs and sneers at the old crusty datasheet logic charts:
Someone here already hit on the most-likely short-term solve for the issues, and it has already been used with success in other devices.
First, in the long-term, the Curtis ICs... at least most of the catalog needs to be reverse engineered with an eye toward copying that sound as precisely as possible, with all the characteristic timbres and "butter" Doug Curtis intended. In many of the CEM designs the specific geometries and locations of the various elements on the dies contribute, in fact were designed to contribute some of that "flavor" everyone born before NKOTB remembers. I'm thinking it's closer to 'possible' than 'impossible'.
For some crazy reason I saw the problem and realized we deal with this kind of issue from time-to-time in my "real job", so I began cornering nerds and engineers in my line of work and forcing long discussions with them. I have gleaned a few positive insights and some hope.
First a few things:
1-Barring any legal ramifications and any hurt feelings, this is possible.. and the short-term solution isn't that difficult if you are so motivated.
2-It's easier than you think. No, not "child's play" but these aren't new components with a great many internals. In many cases some of the CEMs are more "convenient groupings" of oft-used devices for the synth designers.
After manipulating the aforementioned electrical engineers whom I rely on for the deep nitty gritty when wrangling a problem like this, I noticed they were taken aback by the rather simple layout of the CEM ICs... They were correct in realizing that these ICs aren't very complex, and more importantly, the layout, theory and configuration of the "key" CEM ICs were more the reason for the "CEM flavor" than any deep die metallurgy or semiconductor black magic going on. I confirmed my trust in my captive engineering ubernerds' understanding of the relevant ICs by noting that, without yet knowing the nature of the CEM line, they realized simply by my pointing out the odd gate arrays, filter cells and leveling going on that this was some sort of custom line, groupings of oft used, somewhat oddly balanced amps and gates.
I think this can be done, and not only with a high degree of fidelity to the original CEMs, but with a few really great improvements in current use and noise.
The most fun part is torturing the nerds too! This makes for some real "Dilbert" moments with these electrical engineering chip & die people. Nothing sets a GE FADEC engineer's hair on fire better than telling them you kinda want that weird float in stability, an NO we DO NOT consider an "FPGA" an acceptable solution for a synth voice.
(These are the same guys who helped me clone a few Rockman-ish rack items so the Analog Vs SolidState vs. "Rockman" arguments helped again here, and I came to trust their expertise when it comes to nailing then improving upon the performance aspects.)
Sorry, anyway, I only add all that to show a pretty good handle on this problem and with a little help I can offer a path to a solution, if only a valid schematic. People should be able to drop in an equivalent low-run module-board in their nice vintage stuff no?
The temporary solution is to develop a small SMD module, with pins that match the DIP footprints that people can swap in to replace dead CEMs. We have the CEM datasheets and patent writeups despite the die negatives and masks themselves being confirmed lost. With proper care, (and careful study of the precise behavior of the real ICs) it seems possible to create small boards with the functions and behaviors of the CEMs exactly. If you look at the logic layouts and elements you'll realize they're kind of simple at face value. Additionally, such emulations would easily paste into the Oberheim-esque synth I'm looking at making, at the CAD level. With SMD tech in our pockets, there's less of a 1979 manufacturing argument requiring an entire custom package simply because of its' 7 gates and a diode. In 1980 it was a manufacturing edge that saved up to perhaps 20 parts tops. Today that's worth about $7 and 2 seconds of the PickPlace machine's time.
In SMD you can make a passable "555" timer with 8 parts if you were forced to, (Where I went to school you get hit with a breadboard if you can't recall 741s and 4019s like this on the spot) but you get the idea. Naturally this logic has diminishing returns, but in the case of these CEMs they are mostly small arrangements of at best 20 devices which are documented well in most cases on the sheets and in studying the underlying earlier patents. Success is only guaranteed with a careful study of the originals and a fanatical approach to identifying and transferring the aspects of the respective circuit's behaviors to new production designs.
Enroute to a final solution, the work, analysis and tweaking that goes into the build of any SMD dupe modules would also help as a knowledge base toward perhaps bending the right ears and checkbooks and having the line reissued as RoHS complaint so-called "green" modern cross referenced IC. There are arguments pro and con I'd say, but again if a low-run SMD module replaces it to most user's approval why bother?
And last again there's the legal considerations. Even if these patents have expired, what of CEM and the people who created this? My own interest would be to provide acceptable valid module schematics as plans available to all, complete with BOM and footprint infos so people in a pinch could reasonably repair a CEM related product. I don't need the $$ nor do I wish to go into the synth business, just make some great homemade Frankenstein polys and pass the knowledge along.
Nobody gets rich making synths like 1970s TVs but they get people happy!
Thanks for the cool thread and I hope this pushes the discussions further along. I'd hope anyone ay or nay could add their thoughts, I had no ideas anyone else would be that interested beyond a few die hards.
The post is well worth reading BTW. It is an amazing insight into synth chip making. Very Technical, as befitting this thread...and hopefully a way of derailing this thread. Haha.